Running in parallel is much faster but generates an error (sometimes) - it appears to be related to accessing the disk but is not clear (see image). Workaround currently prints tables for each biomarker and tries to save unified table at end. The code appears to get to the last biomarker (which in parallel is not always the same) and fails to write that single biomarker csv table and the final unified one. Workaround is to run in parallel first, and then rerun only on the biomarkers unprocessed
